做法

使用redis的lua脚本功能来限频

在redis中定时刷新系统时间来作为一个全局的时钟

限频脚本：

/**     * 获取令牌的lua脚本     */    public final static String SCRIPT = "local epoch = redis.call(\"hget\",KEYS[1],'Epoch');\n" +            "local currentEpoch = tonumber(ARGV[1]);\n" +            "if(epoch == false and currentEpoch ~= epoch) then\n" +            "  redis.call('hset', KEYS[1], 'Epoch',currentEpoch);\n" +            "  redis.call(\"hset\", KEYS[1],  'Count', tonumber(ARGV[2]));\n" +            "  redis.call('pexpire', KEYS[1], 2000);\n" +            "  return 1;\n" +            "end;   \n" +            "local currentCount = redis.call(\"HINCRBY\", KEYS[1],'Count', tonumber(ARGV[3]));\n" +            "if(currentCount > 0) then  \n" +            "  return 1;\n" +            "else  \n" +            "  return 2;\n" +            "end;";    /**     * 同步系统时间到redis 中的lua脚本     */    public final static String SET_SYSTEM_TIME_SCRIPT = "local clientId = redis.call(\"hget\",KEYS[1],'Client'); \n" +            "if (clientId == false or clientId == ARGV[1]) then \n" +            "   redis.call('hset', KEYS[1], 'Client', ARGV[1]);\n" +            "   redis.call('hset', KEYS[1], 'Time' ,ARGV[2]);\n" +            "   redis.call('pexpire', KEYS[1], tonumber(ARGV[3]));\n" +            "   return 1;\n" +            "end;\n" +            "return 2;";

java调用代码：

List
    
      keys = new ArrayList
     
      ();  keys.add("JRateLimit-Key-" + source);     List
      
        args = new ArrayList
       
        ();   args.add(currentEpoch + "");  //currentEpoch是通过获取全局时钟来做的   args.add(permitsPerEpoch + "");   args.add((0 - permits) + "");    Object result = client.evalsha(LIMIT_LUA_SCRIPT_SHA, keys, args, false);  //result的类型即为lua脚本中的返回值类型，Long
       
      
     
    

策略

通过redis实现全局时钟

通过一次多取几个令牌放到内存中来解决超大调用频率导致的redis请求单机ops极限问题